On the Spatial Statistics of Optical Flow - Analysis, Modeling, and Estimation
Speaker: Stefan Roth , Brown UniversityContact:
Date: November 3 2006
Time: 11:00AM to 12:00AM
Location: Seminar Room D463 (Star)
Host: Gerald Dalley, MIT CSAIL
C. Mario Christoudias, Gerald Dalley, 3-4278, 3-6095, firstname.lastname@example.org, email@example.comRelevant URL:
NO ROOM CHANGE: Apologies for the false alarm. The talk will remain in D463 (Star). It is not being moved to Kiva.
In this talk I will present an analysis of the spatial and temporal statistics of "natural" optical flow fields and a novel flow algorithm that exploits their spatial statistics. Training flow fields are constructed using range images of natural scenes and 3D camera motions recovered from hand-held and car-mounted video sequences. I will present a detailed analysis of optical flow statistics in natural scenes, as well as machine learning methods for learning a Markov random field model of optical flow. The prior probability of a flow field is formulated as a Field-of-Experts model that captures the spatial statistics in overlapping patches and is trained using contrastive divergence. I will compare this new optical flow prior with previous robust priors and show how it can be incorporated into a recent, accurate algorithm for dense optical flow computation. Finally, using experiments with natural and synthetic sequences I will illustrate how the learned optical flow prior quantitatively improves flow accuracy and how it captures the rich spatial structure found in natural scene motion.
See other events that are part of MIT Machine Vision Colloquium 2006/2007
See other events happening in November 2006